About the author(s):
Henrique Garbino is currently a Doctoral Candidate at the Swedish Defence University (2021-present), where he researches the use of landmines and other explosive devices by non-state armed groups. Specialized in explosive ordnance disposal and civil-military coordination, Henrique served in the Brazilian Army (2006-2017) and deployed to border control, law enforcement, and United Nations peace operations. Recently, Henrique worked with the International Committee of the Red Cross as a Weapon Contamination Delegate in Eastern Ukraine (2019-2020; 2022). Henrique has an MSSc in Peace and Conflict Studies from Uppsala University (2019) and a BSc from the Military Academy of Agulhas Negras (2010).
Why are some non-state armed groups more violent than others? Why do some groups resort to inhumane means and methods of war while others restrain from doing so? In trying to answer these questions, a growing number of scholars and practitioners have focused on the drivers of restraint behaviour. However, defining and measuring restraint can be challenging. In this post, Henrique Garbino, Doctoral Candidate at the Swedish Defence University, discusses how we can define and measure restraint focusing on the use of landmines and similar explosive devices by non-state armed groups. This post is based on Henrique’s recent article, “Rebels Against Mines? Legitimacy and Restraint on Landmine Use in the Philippines,” published in Security Studies on 23rd June 2023.
From Ethiopia to Ukraine, from the Maghred to Myanmar, we continue to witness widespread suffering and devastation, leaving civilians either caught in the crossfire or directly targeted by armed groups. However, not all armed groups behave in the same manner. Some show a greater respect for international humanitarian law and work to minimize harm to civilians, while others blatantly flout these rules and engage in atrocities. An ever-growing body of literature has focused on the drivers and dynamics of violence in armed conflicts, but less attention has been paid to restraint behaviour, that is, the self-imposed limitation on the use of violence. Importantly, studying violence is not the same as studying restraint.
Violence and restraint can be seen as opposite sides of the same phenomenon only to the extent that they share the same logic. For example, Stathis Kalyvas’s theory of selective violence posits that armed groups’ motives and capabilities to exercise violence depend on their level of territorial control and, consequently, access to information on sympathisers and defectors. Thus, the theory can explain, at the same time, why groups exercise indiscriminate or selective violence or restrain from doing so. However, drivers of restraint may also follow an independent logic and often present a barrier to the exercise of violence. For instance, Jessica Stanton and Scott Straus show that incentives for violence coexist with incentives for restraint, as the latter moderates specific aspects of the former, such as the strategy and intensity of targeting civilians in armed conflicts.
Thus, focusing on restraint behaviour is important for several reasons. First, it allows us to better explain cases of non-violence or low levels of violence. Second, because restraint can be conceptualised as an integral component in the process of violence—and, indeed, a fundamental component of military strategy, from conventional deterrence to counterinsurgency operations—it provides greater nuance in how we can observe and thus understand the exercise of violence. Finally, from a moral imperative, by better understanding the drivers and dynamics of restraint, it is possible to identify potential pressure points in trying to influence the behaviour of armed groups. As a limit to violence and human suffering, understanding restraint is vital in minimising the human, social, economic, and environmental effects of armed conflict.
The challenges of conceptualising and measuring restraint
The ICRC has, in many ways, spearheaded not only the promotion of restraint behaviour in armed conflicts but also theoretical and methodological reflections on how to conceptualise and measure restraint. For instance, the 2018 ICRC report, “The Roots of Restraint in War,” defines restraint as “behaviour that indicates deliberate actions to limit the use of violence.” In an ensuing series of academic publications based on the report, Brian McQuinn and others further distinguish between genuine restraint and suggestive restraint. While suggestive restraint entails the rhetorical commitment to restraint norms (i.e., “talk the talk”), genuine restraint implies the intention to limit the use of violence (i.e., “walk the walk”). Genuine restraint may stem from both moral and strategic incentives. For example, armed groups may decide against targeting civilians due to genuine respect for humanitarian norms or fear of reprisals, legal indictments, or reputation loss. Finally, mechanical restraint refers to the unintentional limitation of violence, such as the lack of ammunition or impeditive weather conditions (p. 8-9).

The Roots of Restraint in War report also highlighted the challenges in measuring restraint behaviour. To overcome this counterfactual hurdle—that is, the violence that did not happen—scholars have measured restraint against different benchmarks, such as the existing possibilities of inflicting violence, past unrestrained behaviour, different patterns of unrestrained violence by the same armed group; comparable cases of unrestrained violence by other groups; and signs of moral outrage following acts of violence and norms of acceptable behaviour.
Restraint as the application of humanitarian principles
In my research, I define restraint as the self-imposed limitation on the use of violence, despite the existence or the possibility of pursuing such capability. This definition excludes instances of mechanical or suggestive restraint and, at the same time, is agnostic to whether there is an original intention to use violence or not. In other words, restraint can be exercised as the product of a purely rational cost-benefit calculation, where an initial intention to use violence is disregarded due to its disproportional moral or strategic costs. Alternatively, organisational cultures may embed norms of restraint so that armed groups do not even consider using specific means or methods of war. Furthermore, this definition applies to violence in general, including violence targeting combants, and not only violence against civilians or that may cause civilian harm.
In comparison with previous attempts, I argue that measuring restraint against humanitarian norms allows for a broader basis for cross-case comparison, independent from other internal and external contextual references, and befits the existing literature on compliance with and the promotion of humanitarian and human rights norms. Precisely, I measure restraint along four dimensions:
(1) the choice of means and methods of war;
(2) the distinction between combatants and non-combatants;
(3) the proportionality in the use of violence; and
(4) the precaution in the use of violence.
Humanitarian norms restrict the conduct of hostilities in both the means and methods employed by armed groups. First, they establish that the means through which violence is exercised should be restricted. From incendiary bombs to blinding lasers, and from anti-personnel landmines to chemical weapons, some means are perceived to be inherently inhumane and indiscriminate or to cause disproportionate humanitarian and environmental damage that their use should be restricted or completely banned and never used under any circumstance. Second, even when lawful means of warfare are employed, violence can only be exercised following the basic principles of distinction, proportionality, and precaution. Lawful methods of warfare should be discriminate, proportional, and avoid unnecessary suffering. In this sense, violence can only be directed at those participating in the conflict and military objects. To some extent, civilian victimisation—or “collateral damage”—is expected, but it should be proportional to the anticipated military advantage and all possible precautionary measures should be taken to limit the effects of conflict on civilians.
The practical application of these principles will undoubtedly vary widely from context to context. However, I hold they can be operationalised in greater detail for specific types of violence.
Landmines and similar explosive devices
Landmines are often symbolically described as the “perfect soldier” or as the “poor man’s weapon” and thus deemed the weapon of choice of many non-state armed groups. Still, not all groups employ landmines in the same way, as shown in the recent “’From words to deeds” study. While some lay thousands of mines irrespective of collateral damage, other groups directly target civilians, and others attempt to limit civilian victimisation. Some have even unilaterally renounced anti-personnel landmines and actively engaged in mine action. In my recent article, “Rebels Against Mines? Legitimacy and Restraint on Landmine Use in the Philippines,” I attempt to explain this variation of restraint on landmine use.
It follows that restraint on landmine use can be exercised along four main dimensions relative to the practical application of humanitarian norms: the type of device, intended target, location, and precautionary measures. The type of device can be generally classified in relation to the device’s intended effects, i.e. anti-personnel and anti-vehicle, and whether the device is activated by the victim or remotely command-detonated. Victim-activated anti-personnel landmines are more disruptive to civilian life, as inadvertent children and adults can easily activate them. Victim-activated anti-vehicle mines’ impact on civilians is more restricted since they are designed to detonate by the presence, proximity or contact of a vehicle. Finally, command-detonated devices should allow for some degree of distinction between military and civilian targets and the avoidance of collateral damage, given that the indiscriminate effects of explosions—for instance, blast, fragmentation, and heat—can be taken into account.
Concerning the intended target, restraint varies in whether landmines are laid targeting only combatants and military objects or non-combatants and civilian objects in general. Whether it conforms to existing humanitarian norms or not, armed groups might consider unarmed military personnel, police officers, and civilian government supporters as legitimate targets and still differentiate them from other non-combatants, such as ordinary civilians. Civilian victimisation also depends on the location of landmines, that is, whether they are laid in unpopulated areas, populated areas, or directly on or near civilian objects. Finally, restraint on landmine use can be exercised through precautionary measures. Armed groups can reduce the risk of civilian victimisation by issuing early warnings prior to attacks, as well as marking, fencing, and controlling access to minefields or otherwise mined areas.
So what?
What does this conceptualisation tell us about restraint behaviour? Briefly, it allows us to go beyond the dichotomies of use versus non-use of violence or linear scales of violence intensity. Instead, this conceptualisation shows in more detail how restraint and violence are exercised based on core humanitarian norms.
In my article, I find that different factors influence different dimensions of restraint. For example, on the one hand, groups seeking legitimacy from international audiences may stop using anti-personnel mines to comply with the provisions included in Ottawa Convention, a treaty applicable only to States. On the other hand, some groups primarily reliant on local communities may still employ anti-personnel mines but attempt to limit civilian victimisation through marking, fencing, and sharing information on the location of minefields.
I also demonstrate that restraint can only be understood when all dimensions are analysed in conjunction. As mentioned above, some groups may go to great lengths to avoid civilian victimisation even if they do not comply with specific norms. Further, not employing anti-personnel mines is not necessarily an indicator of restraint. Some armed groups may use more selective means, such as command-detonated explosive devices, to target non-combatants directly, thus violating the principle of distinction. Alternatively, they may use such devices in populated areas against legitimate targets in a disproportional way to the expected military advantages.
Finally, similar conceptualisations may prove fruitful in studying specific means and methods of war, such as the use of explosive weapons in populated areas. For example, restraint applied to shelling can be operationalised in relation to the type of munition (i.e., whether high explosive, fragmentation, chemical, incendiary, and so on), intended target, timing (e.g., day or night, week or weekend), intensity (e.g., number of shells, explosive yield of munitions, duration of barrages), and specific precautionary measures (e.g., early warnings, surveillance, fire observation and correction). These findings, in turn, may help practitioners identify engagement strategies and pressure points to encourage armed groups to exercise restraint in particular ways.